See You Next Saturday in Shanghai! AI Compiler Technology Salon Brings Together Big Names From Zhiyuan, ByteDance, Lingchuan Technology, Etc.
In 2023, HyperAI held 4 offline Meet TVM gatherings in Beijing, Shanghai, and Shenzhen.It has gathered more than 1,000 senior practitioners and enthusiasts, and gradually established a rich community ecology. In 2024, HyperAI focused on the field of AI compilers and continued to provide an open communication platform for industry insiders. In July this year, it held the 5th Meet AI Compiler Technology Salon. Now, as 2024 draws to a close, the fierce competition in the large model track has not stopped. What new breakthroughs and new applications are there in AI compilers?
On December 28, HyperAI will hold the 6th Meet AI Compiler Technology Salon in Wujiaochang, Shanghai. The event will be divided into two parts: technology sharing and roundtable discussion.
First of all, we are honored to invite 4 senior experts from Horizon Robotics, Beijing Zhiyuan Artificial Intelligence Research Institute, ByteDance, and Lingchuan Technology to share their application cases and industry observations. In addition, Feng Siyuan, a Ph.D. from Shanghai Jiaotong University and Apache TVM PMC, will host the roundtable session and exchange ideas with the guests from a professional technical perspective.
We have also prepared exquisite gifts and tea breaks for everyone. Please sign up for the event and follow the "HyperAI Super Neuro" public account, and reply "1228 Shanghai" in the background to participate in the lucky draw.The prizes must be collected offline at the event site on December 28th, come and participate~
Event Details
Time: December 28 (Saturday) 13:30-17:45
Location: 2F Lecture Hall, Shanghai Wujiaochang Innovation and Entrepreneurship College (No. 322, Daxue Road, Yangpu District)
Number of people: 200 (Onsite seats are limited, please register as early as possible)
Registration: Click the link to register
Agenda:
Guests and Agenda
Session 1: Guest speakers
Share topic:Transformer model compilation and deployment optimization practice
Content introduction: In recent years, the Transformer model has not only achieved remarkable results in the LLM field, but has also been widely used in the field of computer vision.In the scenario of autonomous driving, the compilation and deployment optimization of the Transformer model is a complex process. The end-to-end model and VLM pose great challenges to the compilation optimization and deployment on the end side. This sharing mainly introduces some of Horizon's practices and experiences in this regard.
Watch this sharing session and you will learn:
1. Challenges in Compilation, Optimization, and Deployment of Visual Transformer Models
2. Horizon's exploration and practice in Transformer model compilation optimization
Share topic:FlagGems: R&D progress and ecological application exploration
Contents:Since the release of version 2.0, with the close collaboration of the Zhiyuan team, partner manufacturers and the Triton China community,FlagGems has achieved significant R&D breakthroughs in the number of operators supported, computing performance, code generation, and runtime mechanisms.At the same time, FlagGems connects the upstream and downstream of the field, and has carried out innovative explorations in large-model training and reasoning adaptation, multi-backend chip integration, and other aspects, further expanding the application scenarios of the operator library in the diversified ecosystem.
Watch this sharing session and you will learn:
1. FlagGems’ latest research and development achievements and technological advancements
2. Innovative practices and experiences in adapting operator libraries to large models
3. Technical solutions and application effects of multi-backend chip integration
4. Progress of Triton China Community and Diversified Ecosystem Construction
Share topic:AI compiler design from the perspective of computation-memory access-communication optimization
Contents:Compilation optimization for AI chips faces optimization challenges from three perspectives: computing, memory access, and communication. This report will share the experience of AI compiler design and optimization in three aspects.From a computing perspective, we share optimizations for customized acceleration units. From a memory access perspective, we share memory access modeling and optimization for AI chips. From a communication perspective, we share experiences and progress in the convergence of computing and communication.
By watching this sharing session, you will learn:
1. A global perspective on AI compiler optimization
2. Learn the experience and skills to fully utilize the computing power of AI chips
3. Understand the future development direction of distributed AI compilers
Share topic:Large model compilation optimization and reasoning acceleration practice based on multi-core architecture NPU
Content introduction: This sharing focuses on the practical exploration of multi-core architecture NPU in large model compilation optimization and reasoning acceleration.By analyzing the computing characteristics of large models such as Transformer, we propose matrix optimization based on tiling strategy, scheduling mechanism of data parallelism and model parallelism, and efficient reasoning method integrating operators and pipeline design. Combined with the parallel characteristics of multi-core architecture, we optimize operator decomposition and memory management, greatly improving reasoning throughput and resource utilization.
Watch this sharing session and you will learn:
1. Architecture of cloud-based large model inference chip
2. Optimization strategy of Attention super operator under Transformer architecture
Session 2 Roundtable Discussion
Roundtable topics:Collaborative development of chips, networks, software, algorithms and ecosystems
Organizers and partners
HyperAI is a leading artificial intelligence and high-performance computing community in China.It aims to help developers and enthusiasts in China's data science and artificial intelligence industry learn, understand and practice by providing various infrastructures such as accelerated data set downloads, online tutorial demonstrations, in-depth interpretation of papers, and top conference calendar integration, and build the future of artificial intelligence with the community. Currently, the official website of Super Neuro has launched thousands of classic and high-quality public data sets and tutorials, and operates the most active AI compiler community in China.
Visit the official website:https://hyper.ai/
OpenBayes Bayesian Computing is a leading high-performance computing service provider in ChinaBy grafting classic software ecosystems and machine learning models onto new-generation heterogeneous chips, it provides industrial enterprises and university scientific research with faster and easier-to-use data science computing products. Its products have been adopted by dozens of large industrial scenarios or leading scientific research institutes.
Visit the official website:https://openbayes.com/
The MLC.AI community was founded in June 2022, and led by Chen Tianqi, the main inventor of Apache TVM and a well-known young scholar in the field of machine learning, the team launched the MLC online course.The key elements and core concepts of machine learning compilation are systematically introduced.
In November 2022, with the joint efforts of MLC.AI community volunteers, the first complete TVM Chinese documentation was launched and successfully hosted on the HyperAI official website, further providing domestic developers interested in machine learning compilation with the basic settings for accessing and learning a new technology - documentation.
MLC Online Courses:https://mlc.ai/
TVM Chinese Documentation:https://tvm.hyper.ai/
Shanghai Wujiaochang Innovation and Entrepreneurship College is a non-profit education and service organization jointly initiated by the Yangpu District Government, famous universities, leading enterprises in the industry, and entrepreneurial service institutions.With the mission of "creating deep connections to make innovation and entrepreneurship more successful", we connect entrepreneurs and enablers, improve the efficiency of allocation of resources for innovation and entrepreneurship, and promote creation.
In November 2022, Wujiaochang Entrepreneurship Academy will officially open the academy's public space, opening up 800 square meters of space to the entrepreneurial crowd. It will also jointly launch the entrepreneurial charity brand "Academy Coffee" with partners such as CUHK Shanghai Center and Shui On Group, and invite KOLs, enterprises and social organizations that have a consensus on entrepreneurial charity services to become "charity partners" to jointly design different forms of ways for entrepreneurs to gather.So far, "College Coffee" has held more than 140 dual-entrepreneurship themed activities in various forms, bringing together more than 3,000 participants.
Event Support
